A Principled Approach to Analyze Expressiveness and Accuracy of Graph Neural Networks - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2019

A Principled Approach to Analyze Expressiveness and Accuracy of Graph Neural Networks

Résumé

Graph neural networks (GNNs) have known an increasing success recently, with many GNN variants achieving state-of-the-art results on node and graph classification tasks. The proposed GNNs, however, often implement complex node and graph embedding schemes, which makes challenging to explain their performance. In this paper, we investigate the link between a GNN's expressiveness, that is, its ability to map different graphs to different representations, and its generalization performance in a graph classification setting. In particular , we propose a principled experimental procedure where we (i) define a practical measure for expressiveness, (ii) introduce an expressiveness-based loss function that we use to train a simple yet practical GNN that is permutation-invariant, (iii) illustrate our procedure on benchmark graph classification problems and on an original real-world application. Our results reveal that expressiveness alone does not guarantee a better performance, and that a powerful GNN should be able to produce graph representations that are well separated with respect to the class of the corresponding graphs.
Fichier principal
Vignette du fichier
Expressiveness-analysis-GNNs.pdf (232.69 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02378826 , version 1 (21-11-2019)
hal-02378826 , version 2 (25-11-2019)

Identifiants

  • HAL Id : hal-02378826 , version 1

Citer

Asma Atamna, Nataliya Sokolovska, Jean-Claude Crivello. A Principled Approach to Analyze Expressiveness and Accuracy of Graph Neural Networks. 2019. ⟨hal-02378826v1⟩
183 Consultations
291 Téléchargements

Partager

Gmail Facebook X LinkedIn More